06-15, 16:30–17:10 (Europe/London), Warwick
functime is a modern time-series forecasting library to generate predictions for thousands of time series at once, while never leaving your laptop. Thanks to Polars' powerful query engine, feature extraction and cross-validation are 1-2 orders of magnitude faster. Plus, functime offers a best-of-the-class set of diagnostic tools to further streamline your workflow.
In this talk, we'll learn how to use functime to analyse your model and generate blazingly fast prediction intervals using EnBPI, a state-of-the-art conformal prediction framework that is also available in other popular Python packages.
functime is a modern time-series forecasting library, built on top of Polars to harness its powerful query engine and provide a new take on time-series analysis. Generating predictions for large panel datasets can be a costly and time-consuming endeavour: univariate forecasting is the work of a tailor, while model deployment and monitoring usually requires more sophisticate engineering, and sometimes even distributed systems.
functime addresses these challenges by leveraging Polars' query engine, which can trivially parallelise time series forecasting tasks. As a result, functime enables practitioners to generate predictions for thousands of time series simultaneously, all from the comfort of their laptops. This efficient forecasting workflow not only leads to dramatic improvements in speed but also enhances the overall development experience, thanks to a powerful set of diagnostic tools, an sklearn-inspired API and lightweight abstractions to work on multiple time series at once.
One of the key features of functime is first-class support for uncertainty estimation through EnBPI (Ensemble Batch Prediction Intervals), a state-of-the-art conformal prediction framework. Conformal predictions offer an assumption-free way to generate prediction intervals, and EnBPI is even faster, as it does not require neither data-splitting nor training multiple ensemble estimators.
This talk is meant for forecasting practitioners of all levels, as well as data scientists interested in uncertainty estimation. We will explain conformal predictions and how EnBPI is implemented, and showcase how functime abstracts away this complexity to streamline the forecasting workflow.
🚩 Talk outline
• minutes 0-5. Problem setting: Challenges in large-scale time series forecasting.
• minutes 5-10. Introduction to functime: how it leverages Polars for exceptional performance.
• minutes 10-20. EnBPI and conformal predictions: generating prediction intervals for time series forecasting.
• minutes 20-35. Show me the code: A simple forecasting workflow with functime with uncertainty quantification.
• minutes 35-40. Conclusion + Q&A.
No previous knowledge expected
ML Engineer